Defining Cognitive Boundaries for AI in Education: A Governance Framework for Schools under the EU AI Act (RAGA Model)

Artificial intelligence (AI) is being rapidly integrated into educational environments across Europe, creating new challenges for institutional governance. While the European Union Artificial Intelligence Act (Regulation (EU) 2024/1689) establishes a risk-based regulatory framework, schools are increasingly positioned as deployers of AI systems and must therefore ensure responsible implementation in practice. However, existing school-level AI policies often remain high-level, ethical, and descriptive, failing to provide operational clarity for classroom decision-making. This paper proposes a structured implementation framework for AI governance in schools, grounded in the principle that AI use should be determined by the cognitive demands of learning tasks rather than by technological capability. Building on this premise, the paper introduces the Responsible AI Governance Architecture (RAGA), a five-layer model integrating policy positioning, risk classification, pedagogical boundaries, teacher capability, and institutional oversight. In addition, the paper presents a practical 30-day implementation roadmap designed to support schools in transitioning from policy development to operational governance. Through applied examples, including AI use in assessment, student writing, and ideation, the framework demonstrates how institutions can align regulatory compliance with pedagogical integrity. The paper argues that effective AI governance in education requires a shift from abstract ethical guidance to decision-oriented frameworks that define where AI must be limited to preserve meaningful learning. The RAGA model offers a structured approach for schools seeking to operationalise AI governance in alignment with European regulatory expectations while maintaining the integrity of student cognition and assessment.